I don't think that all the measurements in the world will give a correct frame size. If I'm riding a road bike hard, I'm comfortable in a much lower position, with a longer reach than I am if I'm riding at a low power output. My measurements haven't changed. My road bike, however, has developed a longer reach over the period I've owned it - I used to ride it with bars and saddle level, now they're considerably lower and while the stem's a touch shorter, the bar angle puts the hoods further away.
To me, frame size depends on torso length, (EDIT: also arm length, with an appropriate bend in the elbow) the rider's flexibility and core strength, their power output, and their weight. That's why I believe in a "magic number" found through test-riding. I suppose if one could quantify flexibility and core strength, measure the rider's weight and weight distribution from the hip up only, and model all of that, that set of equations might be an okay predictive model for sizing frames.
